Share this post:
Today’s business leaders know that AI is playing a critical role in transforming customer service. Customers expect answers in real time, on the channel they want and at the time they want.
Watson conversational AI makes it easy for your customers to communicate with your organization by having a simple conversation and for your business to respond intelligently across multiple touchpoints. Intelligent end-to-end customer service allows you to harvest and use conversational data across channels to understand customers in near-real time, driving ever greater innovation, personalization, and customer engagement.
Conversational AI is solving these 3 customer services challenges
What if you could understand who your customers are, and what they want, before they dial your support number or open a web chat? And what if your customer support teams across channels had access to the same data, in context, when they interacted with customers? Finally, what if you had a way to automate those most common, simple support calls and could put the best answers to complex customer questions into live agents’ hands with speed and accuracy?
Businesses in industries as diverse as insurance, telecom, legal, marketing, and energy are applying Watson conversational AI to address exactly these customer service challenges with astonishing results.
How does conversational AI deliver these kinds of results?
Conversational AI, or virtual assistants, are not the same as chatbots, which are single-mode and linear, responding to specifically worded questions based on a simple Q&A script, like an automated FAQ. A chatbot can’t understand context or conversational flows that deviate from the scripts they’ve learned.
Virtual assistants on the other hand communicate more naturally and can understand context. Through the use of machine learning and core conversational AI concepts they can be trained to understand idioms, slang, common abbreviations, and phrases that deviate from the script and that clients might use when talking with a live agent. They can also manage multiple nodes and determine which among a number of possible responses would be most appropriate, including handing off to a live agent for interactions it can’t recognize or resolve. Here’s a simple interaction that illustrates the difference.
Samantha wants an unlimited cell phone plan. She’s seen competitors’ ads and checks her provider’s site. A customer service chat window pops up.
Account Services Bot
Hello, Samantha. This is the Account Services Bot. How may I assist you?
Hey, I just saw an ad from a competitor offering uncapped data plans for less than what I pay. Any chance you guys would match that?
Account Services Bot
Sorry, Samantha. We do not offer uncapped data plans. I apologize for any inconvenience this may cause. Please try us again after 9 a.m. Pacific Standard Time, and a live agent will be more than happy to assist you further. Was there anything else I can assist you with?
With a virtual assistant the experience could be quite different.
Account Services Virtual Assistant
Yes, Samantha, we do offer an unlimited data plan at nearly the same rate as your current plan … then the virtual assistant could continue to interact with Samantha and discern which conversational dialog node would be most appropriate:
- Review the available plans
- Guide her through changing her plan
- Offer her a pre-approved ‘loyalty’ discount
- Offer to transfer her to a live agent
Modern conversational AI platforms like Watson allow for contextual bias, can understand a broader understanding of human language inputs including regional dialects, and can even assist customers to complete transactions based on their advanced ability to guide a customer through a conversation.
6 steps to effective conversational design
Conversational design is a process of planning for your customers’ needs and crafting the flow of typical conversations, building dialogue, writing responses, and creating opportunities for your conversational AI assistant to provide value, learn, and grow. There are six steps.
- Collect questions and understand utterances
What do your customers want to know? The best way to figure that out is to read what they’re already asking and look at what they’re already doing in their interactions with different parts of your organization. How do they speak? What slang or acronyms do they use? Identify enough examples to capture the typical ways your customers express key concepts that are important to your specific use case.
- Map your ground truth and identify intent clusters
Mapping ground truth allows you to clearly map actions to intents, or concepts your customers are likely to reference when engaging with your conversational AI solution. Group phrases, or utterances, with the same meaning together. Have at least ten different utterances for each intent. For example, “I’m frustrated; I haven’t been able to login in to the online billing system,” and “I can’t get into my account, can you help me” could both be in the Password Reset intent cluster, and trigger the same action.
- Design the dialogue
Designing your AI assistant’s dialogue involves customizing the words, phrases, and Q&A that make up the user experience of personality, positioning and proactivity. Design factors include tone, purpose, and role in the conversation.
- Crafting the conversational flow
For each conversation, determine the steps you want the customer to take and the conversational flow to lead them in that direction. To do this, start at the outcome you want, and then work your way forward to the very first welcome by the conversational AI.
- Designing responses
Design concise, engaging responses that will make your virtual assistant’s interactions feel natural by adjusting for tone, restating intents, and introducing variation in the conversation.
- Continuous learning
Build in continuous learning and monitor customer usage patterns to ensure that your virtual assistant gains capabilities and stays up-to-date on product, industry and other changes that impact customer service.
Applying advanced conversational AI concepts to scale and enhance your solutions
We’ve covered the basics of conversational AI. Now it’s time to look at three extremely valuable, advanced conversational AI concepts that can radically enhance your solution: Natural conversations, integrations and analytics.
Natural conversation features in Watson Assistant can be used to help your conversational AI solution better understand customers’ intents and take appropriate action, whether offering a one-off response, collecting more information from the customer, or digressing to address a new or related topic raised by the customer.
The contextual entities feature allows your assistant to detect entities based on context. It’s impossible to define every possible entity, or thing, that a customer might mention, fortunately, with contextual entities we can help the system understand the context in which any type of entity might appear. For example, rather than programming every possible type of cuisine into your assistant, you could train it to recognize the context where a type of cuisine might show up in a conversation.
Slots and digressions are other natural conversations features available in Watson Assistant and used to improve a conversational AI’s ability to understand intents based on context.
The integrations feature in Watson Assistant helps your conversational AI solution cost-effectively work with a wide variety of front- and back-end systems – including phone calls, websites, social media conversations, instant messenger chat sessions, SMS text messaging, CRM systems, IoT devices and more.
With the analytics feature in Watson Assistant, Watson collects analytics on what type of questions are being asked in the thousands of calls it conducts each day, and how they’re answered, allowing you to understand how your customers and virtual assistants are interacting with each other and identify opportunities to improve performance and gain customer intelligence.
Not that long ago, executing advanced conversational AI concepts like these would have taken hundreds of thousands of lines of expensive bespoke code, if they were possible at all. And before that, the data and understanding hidden within it would have simply been inaccessible.
Today, Watson Assistant handles much of this work, dramatically lowering the barrier to adding conversational AI to your business.
Explore our virtual Masterclass series on customer service and AI. Watch presentations with leading experts in AI including Rob High, Vice President and IBM Fellow, CTO IBM Edge Computing, with Lakisha Hall, Director, Watson Expert & Delivery Services, and the Autodesk team responsible for AVA, their conversational solution that helped reduced resolution times by 99%.